Sizing the common mode choke for output of a PWM inverter

You need to know what you are trying to block with the choke. The ideal common mode choke has a very high and resistive impedance at the troublesome frequencies.

In general, you want the highest mu you can find. Saturation should not be an issue because the currents should match.

If you can afford to do so, you may want to put a common mode choke on both the input and the output. The ground of the converter section should have only one route by which its clatter can get to ground. That route should be very low impedance.

The common mode choke is creating a high impedance in one route. Unless there is a lower impedance route, the common mode choke is less effective.

Simple answer: One approach to this problem is to design your circuit so that your CM filter requirements are minimal, then choose CM filter components that aren't so big that their parasitics end up ruining their filter performance.

Details:

1) First, do everything you can do to minimize capacitance from high dv/dt nodes to anything relating to input or output wires, components, connectors. 2) Provide designed minimum loop area paths for any remaining parasitic currents, e.g., snubbers and shunt capacitors. For example, for parasitic currents flowing through primary to secondary capacitances, electrostatically shield pri to sec and/or add shunt capacitance such that the inevitable return current associated with the parasitic current can flow in a capacitor placed very near to the transformer. In order to design these paths, you'll need to locate all high dv/dt elements and at least imagine where their parasitic currents will flow in complete circuit, returning to the source. I = C dv/dt, so you cannot 'block' the current, only reduce it by reducing C or dv/dt, or manage it by shunting it properly. Quite often, large currents flow to heatsinks of switching transistors and rectifiers, and the return paths for these currents should be considered and controlled. Designers sometimes use heatsink spacers to reduce parasitic C or even an intervening shield between the semiconductor and the heatsink. 3) Bring differential EMI to acceptable levels using shunt capacitors and series inductance (if necessary). Some series L may be available from the leakage inductance of your CM filter (see below). 4) All of the above measures will also reduce CM EMI. So, after the above steps, evaluate the need for CM filtering. Important considerations usually include cost, size, and weight. A common mistake is to choose the largest possible components, thinking that they will provide the best filtering. However, the parasitic capacitance of the CM filter components to other circuit components is often the determining factor in filter performance. Stray magnetic coupling is also sometimes a factor. In other words, your nice big CM transformer (aka CM choke) can have a lot of parasitic capacitance to that MOSFET on heatsink that might be just a few cm away, and you suddenly have 100s of microamps of parasitic current flowing directly into (one side of) your CM choke. For the same reason, the placement and orientation of the filter components are critical. Complete CM filters in metal enclosures or inside your own shielding components will generally have superior performance to collections of filter components unshielded from the switching elements. 5) Quite small CM chokes an be very effective, particularly in combination with shunt capacitors. The main limitation tends to be wire size: you need to use a wire size large enough so that heat is not a problem. Obviously, a larger core makes it possible to achieve more inductance for a larger wire size, but beware the parasitic C. The CM choke will also have its own parasitic capacitance from turn to turn, which provides a shunt path for EMI. You also need to insure that any differential currents flowing in the choke are quite small, since saturation is otherwise a possibility. 6) Many CM chokes are deliberately wound so that they develop a small amount of leakage inductance that can help with DM EMI filtering. This is done by putting the windings on opposite sides of a toroid or legs of an E-core.

Paul Mathews

Paul, all great stuff. The OP said DC/AC converter so the input DC return may possibly be tied to the chassis then a input CM choke is not needed. WSU? Harry

Depends greatly on how the power is brought into the enclosure and how the power connections connect to the outside world. If any appreciable amount of unshielded input power conductor is exposed to dv/dt, the resulting parasitic currents/voltages can exit the enclosure and radiate from cabling. However, low voltage switching, if that's what the OP is doing, does have the advantage of smaller dv in the first place. Paul Mathews

Thanks Paul. The reason why I'm asking this question is because I saw there is a CM choke used at the output stage of a 230VAC 300W inverter I bought. In the spirit of "reverse engineering" I thought I should steal the same idea and follow the same in my circuit design, although I haven't yet figured out clearly the cause and principal of common mode current ( I've read a bit and that it's somehow due to transformer leakage capacitance and the cap between heatsink and something ).

My design follows the common architecture of most 12VDC-230VAC inverter: DC-DC (full bridge converter) then DC-AC (PIC driving H- bridge using IRF840). I've already made both parts of circuit working on breadboard (although haven't really loaded it yet since the breadboard can't handle any huge current on low voltage side) and is now entering the stage of PCB design. But before the circuit is finalized I wish to add the extra bit the CM choke in order to make my thesis supervisor happier. If the choice of inductance value happens to be a matter of trial and error as what Eeyore said, what conservative value would you guys suggest me to use, so that I can order from supplier straight away and start with the layout design?

THX

Or.......In other words, how much the impedance CM choke appearing to switching frequency (39kHz) should I design

On Aug 24, 1:37 am, snipped-for-privacy@hotmail.com wrote:

It depends on the nature of the hash on the output, some of which is invariably CM (for reasons described earlier, i.e., parasitics). For a DCDC converter, there is often energy in the 10s to 100s of MHz that results from rectifier turn-off transients. It's usually best to snub the rectifiers for this, which takes care of both CM and DM. However, any remaining CM requires a relatively small inductance due the high frequency. Obviously, the SRF of the inductor must be quite high, so single layer windings are the rule. For DCAC, there are no secondary rectifiers, and the transients on the secondary are a combination of primary-side transients seen through the turns-ratio of the transformer and transients arising from parasitic coupling from the primary side. Begin by minimizing all sources. Then, common mode couple a wide BW oscilloscope or spectrum analyzer to the output and see what's left. (One way to do this is to pass all output conductors through a wideband current probe.) If you don't have any means of probing to see what's there, guess that output CM EMI will have same spectrum as primary DM switching transients, which you should be able to observe. Suppose that this means you're trying to filter a broad spectrum around 10 MHz. You're going to want a few Kohms of impedance in CM, so 50 uH or so will get you there. Find a ferrite toroid that fits your housing and budget and see if a single-layer widely-spaced double winding of adequately sized wire will provide that much inductance. If not, go for a larger core. Paul Mathews